Issue 22: Networking and 5G foundational for AI
Sharmilli Ghosh
Product Management | GTM | ISV & SI Partnerships | Startup Founder | Board Member | Investor |
In my most recent role at Microsoft, we were getting ready for AI to disrupt the Telecommunications industry. While the journey and speed of AI integration is unique for each operator, it is surely driving innovation, efficiency, better customer experience and generating new revenue streams and business value.
The global generative AI in telecom market size is expected to hit around USD 4,883.78 million by 2032, growing at a CAGR of 41.59% during the forecast period from 2023 to 2032. This is one of highest CAGR's seen in any industry. Rapid innovation in Chatbots, managing enormous volumes of customer queries, to predictive network optimizations and fraud detection, among other scenarios, is driving this growth.
Today's article is brought by colleague and domain expert Shubh Agarwal as a guest author.
Introduction
The disruption of AI is dominating your reading or listening lists and influencing your investment paradigms. My host AI expert blogger has been covering a wide range of AI use cases. IT leaders across automotive, financial, healthcare, education, e-commerce, consumer brands are busy investigating a plethora of applications to determine implications to their customers, the company and even their own careers. Jury is out in many minds, but the innovations and the money flow are both white hot driving early adoption in accelerated time frames. I am a firm believer in AI disruption, simply based on its adoption by the next generation, which is always a compelling success indicator.
The mobile industry has seen its heydays, but took a back seat to other disruptive waves of innovations, while continuing to create massive value. The hard fact is that the trillions of dollar value created in e-Commerce, smart phones, social and cloud industries have all anchored their success on global mobile networks serving as the critical arteries in our age of information. The Telcos have been ahead of the technology curve and at a rapid pace and efficiency, as price per bit, a global measure for the value of mobile data, has been declining for the 5.6 billion connected consumers.??
It is now the time for modern mobile networks, aka 5G, to make AI successful. Yes, AI absolutely needs networking and? in many cases 5G to be relevant.? This blog categorizes the use cases in context of connectivity requirements.
Note: the term "5G" is used widely by telecom operators for 4G LTE networks enhanced by 5G NR radios, termed as? 5G Non Standalone.? While this blog focuses on the importance of mobile connectivity in general, the next generation 5G Standalone networks, which are sparsely deployed, provide the enhanced capabilities required for deploying AI applications at scale.
Key technologies for AI applications
AI is a culmination of multiple technologies as key enablers for complex systems. They require not only ML technologies, but also trustworthy data sensors and sources, reliable connectivity,? appropriate data conditioning processes, responsible governance frameworks, and a balance between human and machine interactions.? Enterprises must evolve into a comprehensive systems engineering and ecosystem mindset to develop their appropriate applications based on the target business value.??
Here is a summary of key technologies required for AI applications.?
Foundational: Big Data Technologies, Data Storage and Databases, Edge Computing Technologies, IoT Devices, Graphical Processing Units (GPUs) and Hardware Accelerators, LAN, WAN & Mobile Networking.??
Core AI: Machine Learning Frameworks, Deep Learning Technologies, LLM and Generative AI techniques and models
Application Development: Cloud Platforms, Programming Languages, Data Processing and Analytics Tools, APIs and SDKs, DevOps Tools, Security Frameworks
AI Use Cases by Connectivity Requirements
AI is already around us and is revolutionizing many industries, including energy, consumer products and services, automotive, financial services, national security, healthcare, and advertising.? We are also witnessing the rapid deployment of new and open LLM models with Generative AI due to its inherent adoption simplicity. ? This is an attempt to categorize the use cases in terms of connectivity requirements, especially the need for 5G.
Asynchronous AI: This category covers a broad range of use cases already around us including Marketing or Shopping? engines on our mobile and PC.? ? Evolution of existing Enterprise Data Science and ML techniques are benefiting from AI for enhanced and rapid business decisions, such as dynamic pricing or credit approvals.???
These use cases rely on global networks to collect data, process them in your PC, enterprise data centers or massive cloud? farms, and independently interact with the users based on processed outcomes.? While today’ broadband networks deliver the current demand, 5G will allow broader access through satellite coverage, more capacity and connectivity for more IoT data for Smart City or Agriculture use cases.? Machine learning algorithms rely heavily on data to make accurate predictions and improve their performance. With 5G's lightning-fast network capabilities, these algorithms can access and process data in real-time, enabling faster and more efficient learning.
Near Real-Time, Generative AI: Widely used Search and Social networks rely on continuous asynchronous discovery and learning of data in the backend, which is used to provide a near-real-time engagement for the end users with AI content offered on-demand. ? Connectivity is foundational for data collection, ingestion and distribution.
New use cases enabled by LLM and Generative AI also rely on massive amounts of processing independent of the user interaction, which can handle conversational level latency in the range of 200 ms or higher.? Autonomous vehicles rely on local AI processing but will rely on public networks to share and get environment data or update models for enhanced safety. ? AR/VR is also steadily gaining traction with Meta, Apple and Magic Leap in the game, using onboard processing, assisted by online data.??
As these use cases become prevalent and richer with images and video, the need for reliable connectivity, bandwidth and capacity on public networks becomes relevant, where 5G slicing and dynamic control will be key to power customized experiences per use case.??
Synchronous or Real-time AI: Robotics are gaining fresh attention with humanoid developments like Tesla Optimus and Figure, as an evolution to a strong market for Industrial Robotics.? Video analytics adoption is gaining traction evolving from security to asset tracking, user engagement and safety applications.??
All of them are inherently mobile devices which require highly reliable coverage, bandwidth and ultra low latency that can only be achieved by 5G. ? Additional use cases such as large scale industries, complex environments such as ports and airports that leverage massive amounts and types of sensors with low amounts of data, will also rely on high capacity 5G networks.
领英推荐
The case for 5G in AI
One of the most significant impacts of 5G in AI applications is its ability to unlock the full potential of real-time data analytics. With its enhanced data speed and capacity, 5G enables AI systems to process vast amounts of data at lightning-fast speeds, allowing for quicker decision-making and more accurate predictions.
The role of 5G networks in AI applications is significant and multifaceted, offering various benefits and new opportunities:
Enhanced Data Speed: 5G networks provide much higher data speeds and capacity compared to previous generations. This enables AI applications to process and analyze large volumes of data in real-time, essential for video analytics, autonomous vehicles, and smart city applications.
Reduced Latency: 5G offers significantly lower latency, which is the delay before a transfer of data begins following an instruction. This is critical for AI applications requiring real-time decision-making, such as in robotics, gaming, and augmented reality.
Improved Capacity: 5G's improved connectivity allows more devices to be connected simultaneously. This is crucial for the Internet of Things (IoT), where numerous sensors and devices communicate and share data, often analyzed by AI algorithms for efficient operations, predictive maintenance, and personalized user experiences.
Enhanced Reliability: 5G networks offer more reliable connections with dedicated licensed spectrum, coverage overlap and 99.999% reliable architectures.? This is crucial for AI applications in critical infrastructure and services, ensuring that AI-driven processes are consistent and uninterrupted.
Flexible Edge Deployment: 5G networks are designed to work seamlessly with edge computing and even device-to-device communication. This means data can be processed closer to where it's generated rather than being sent to distant servers. AI applications like drone swarms, autonomous driving or emergency response systems will benefit with faster processing, lower response times, reliability and cost efficiency.
Global Reach and Accessibility: With broader coverage, 5G can extend the reach of AI applications to more rural and remote areas, making the benefits of AI more accessible across different regions and consumers.
Tech Note:? Latest Wi-Fi advancements use similar radio technology as 5G to deliver high bandwidth and offer a very low cost solution for many IoT use cases.? However, it is limited to unreliable unlicensed spectrum, and lacks the mobility, global coverage, end-to-end control or enhanced security frameworks of 5G.??
In conclusion, the impact of 5G in AI applications is transformative. From enabling real-time data analytics to improving responsiveness and efficiency, the combination of 5G and AI opens up a world of possibilities for innovation and advancement in numerous sectors. Stay tuned for the next section, where we will delve deeper into specific industry applications of this revolutionary duo.
References
Coming up Next